Optimize vector clock merge/compare/increment and binary serialization

[dexcompiler]

This PR improves VectorClock performance by reducing allocations and avoiding repeated lookups in the most common operations (Compare(VectorClock), Merge(VectorClock), Increment(ushort)) and by using BinaryPrimitives for big-endian binary encoding/decoding. No behavior or wire-format changes are intended.
Changes:

• Compare(VectorClock): replaced HashSet + repeated Get( ) calls with a single linear merge-style walk over both sorted node-id arrays (allocation-free, O(n+m)).
• Merge(VectorClock): removed List<T> usage and ToArray( ) copies; now merges directly into pre-allocated arrays and trims to actual size.
• Increment(ushort): when incrementing an existing node, share the immutable node-id array and copy only the counters array; small copy optimizations for insert case.
• WriteTo(Span<byte>) / ReadFrom(ReadOnlySpan<byte>): use System.Buffers.Binary.BinaryPrimitives for big-endian reads/writes (same on-wire format).
• Docs: updated complexity claim in the VectorClock XML comment from O(n) to O(n+m).
• Repo guidance: updated .github/copilot-instructions.md to prefer mathematically precise complexity docs during refactors while preserving semantics/wire format.

[dexcompiler]

@copilot review

[Copilot]

@dexcompiler I've opened a new pull request, #33, to work on those changes. Once the pull request is ready, I'll request review from you.